Change the typedefs in lnd ko2iblnd to proper structures.
Several other style changes to fix checkpatch issues with
code impacted by typedef change.

Test-Parameters: trivial

Change-Id: I55e9c91e392dee804802153bd609afc858a3591b
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/32802


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Sonia Sharma <sharmaso@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

This enables server build for Ubuntu 18.04 LTS, the ldiskfs
patches are based on Gael's 4.12 support,
they apply to kernel versions 4.15.0-20.21 to 4.15.0-23.25

There's also a small fix to make dpkg happy when installing
lustre packages which requires lustre-client-utils.

Test-Parameters: clientdistro=ubuntu1604 trivial
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Signed-off-by: Gael Delbary <gael.delbary@cea.fr>
Change-Id: I65e1a5ee0d17115f23ba071ff1ab23b4fb22e78f
Reviewed-on: https://review.whamcloud.com/32613


Tested-by: Jenkins
Reviewed-by: Quentin Bouget <quentin.bouget@cea.fr>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Yang Sheng <ys@whamcloud.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

We're allowed to start at a non-aligned page offset in the first
fragment and end at a non-aligned page offset in the last fragment.

When checking the iovec exclude both of the first and last fragments
from the tx_gaps check.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I8a9231db7db404a5d5a6294ff263c1bd2ac28e6c
Reviewed-on: https://review.whamcloud.com/32586


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

Reverse bte rdma transactions stats were being
incremented by kgnilnd_admin_addref() which asserts when the value
goes negative. These stats should be incremented with atomic_inc
instead.

Test-Parameters: trivial
Cray-bug-id: LUS-5940
Signed-off-by: Chuck Fossen <chuckf@cray.com>
Change-Id: I06426bc078cc76f14c7b3efb5f3ceb71054c2d09
Reviewed-on: https://review.whamcloud.com/32184


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

In lnet_ipif_enumerate(), remove the CERROR() that prints each device.

Test-Parameters: trivial
Signed-off-by: John L. Hammond <john.hammond@intel.com>
Change-Id: Ida8d1636e9e608087205defabda865f930fd38a1
Reviewed-on: https://review.whamcloud.com/32560


Tested-by: Jenkins
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

This commit is supposed to fix armv7 Lustre client
build, mostly 64-bit division related changes.

Change-Id: I93d83d577351c1a1053e39a162cb1e85fc4e8aa3
Signed-off-by: Andrew Perepechko <c17827@cray.com>
Reviewed-on: https://review.whamcloud.com/32194


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Alexander Zarochentsev <c17826@cray.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The ioctl request was designed to get all needed information
through socket from usermode. But the same patterns with tricks
was used at kernel by lib-socket..
The patch changes this behavior from socket to kernel socket.
For libcfs_sock_ioctl tricks with usermode call removed,
kernel_sock_ioctl is used instead. But this call handle SIOCGIFADDR
and SIOCGIFNETMASK. For SIOCGIFFLAGS we take device flag directly.
Function libcfs_ipif_enumerate() which are used SIOCGIFCONF command
totally rewriten to get device names without ioctl requests.

Signed-off-by: Alexander Boyko <c17825@cray.com>
Change-Id: Idf0da800a49dbefa419fc5fecaa9ee1bd4d85327
Reviewed-on: https://review.whamcloud.com/14179


Tested-by: Jenkins
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Linux kernel 4.15 replaced setup_timer() with the new
interface timer_setup().
Introduce cfs wrappers to handle the API changes.

Linux-commit: e99e88a9d2b067465adaa9c111ada99a041bef9a

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Change-Id: Ib79495f9ab7e955d6f72f1e9390cec0e23e2d641
Reviewed-on: https://review.whamcloud.com/31790


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: Jinshan Xiong <jinshan.xiong@gmail.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

In the lustre code various small bits are left of the libcfs time
wrappers. Remove all the remaining wrappers except for the inline
function cfs_time_seconds(). For cfs_time_seconds() we have to
move to nsec_to_jiffies() since msecs_to_jiffies() has become an
inline function in jiffies.h which means HZ can be different on
the installation node verses what the target node is configured
for. Since nsec_to_jiffies() is a normal built in kernel function
we can avoid the LU-5443 issues. For cfs_duration_secs() we use
the internal kernel function jiffies_to_msec() since only the
newest kernels have jiffies_to_nsec() and that function is just
a wrapper around jiffies_to_msec(). All the reset of the libcfs
time abstractions are gone.

Change-Id: I166973071304f1d55a15b1e21fcfbe434ff58199
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/31068


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Ben Evans <bevans@cray.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Quiet warnings in configure checks to avoid confusion when debugging
autoconf/configure scripts, and to avoid potential errors during
detection.

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Change-Id: I2916a120b57fb9fa529bf7050cf65511233ebbe5
Reviewed-on: https://review.whamcloud.com/31904


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Allow peers to be added using ip2nets syntax, from
both command line and YAML block.

Command line example:
lnetctl peer add --ip2nets 10.10.10.[3-6,9]@tcp
lnetctl peer del --ip2nets 10.10.10.[3-6,9]@tcp

YAML example:
peer:
    ip2nets:
      - nid: 30.10.10.[3-8]@tcp
      - nid: 40.10.10.[9-40]@tcp

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I904be8d496ad2be277c3d21dc7f72cbc7ed02b50
Reviewed-on: https://review.whamcloud.com/31786


Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Support the ip2nets syntax for route gateway configuration. Only support
a maximum of 128 gateways per configuration command. This upper limit is
to prevent a large number of routes to be configured by mistake. This
feature is available from both command line and YAML interfaces.

Command line examples:
lnetctl route add --net tcp1 --gateway 12.3.4.[2-6]@tcp
lnetctl route del --net tcp1 --gateway 12.3.4.[2-6]@tcp

YAML examples:
route:
    - net: tcp1
      gateway: 12.3.4.[2-20]@tcp

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I0d8a0f94cce7140602a64f13f0401ef209f3ca57
Reviewed-on: https://review.whamcloud.com/30511


Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

From Linux 4.15 kernel the set/get function prototypes
for module_param_call has changed. This triggers compile
error:
  CC [M]  lustre-release/lnet/lnet/api-ni.o
In file included from lustre-release/lnet/lnet/api-ni.c:38:0:
include/linux/moduleparam.h:232:22: error:
initialization from incompatible pointer type [-Werror]
  static const struct kernel_param_ops __param_ops_##name =
We can switch to module_param and cfs_kernel_param_arg_t since
they are already available in libcfs.

Linux-commit: b2f270e8747387335d80428c576118e7d87f69cc

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Change-Id: Ie6f0f3a31fd22904e03e8300f45f0a8684265abd
Reviewed-on: https://review.whamcloud.com/31788


Tested-by: Jenkins
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The default behavior of LNet is to always use 256 pages which is
LNET_MAX_IOV and that LNET_MAX_PAYLOAD is always one megabyte.
This assumes pages are always 4K in size which is not the case.
This cause bulk I/O errors when using platforms like PowerPC or
ARM which tend to use 64K pages. This is resolved by first making
LNET_MAX_PAYLOAD always one megabyte since this is what the
configuring sets it too by default and no one ever changes it.
In theory it could set it to as high as 16MB but that will cause
the I/O errors since the ptlrpc layer expects the packets to be
always 1 megabyte in size. Also it would be better to make the
maximum payload a per network setup configurations instead of for
everything. Second we make LNET_MAX_IOV equal to LNET_MAX_PAYLOAD
divided by the PAGE_SIZE. This way packets will always be the
LNET_MAX_PAYLOAD in size but the number of pages used,
LNET_MAX_IOV will vary depending on the platform it is creating
packets on.

Change-Id: Ie1dcdb195e68b44e2fa2d9b24715216d8aca4c65
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/31559


Tested-by: Jenkins
Reviewed-by: Wang Shilong <wshilong@ddn.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Ruth Klundt <rklundt@sandia.gov>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

In cfs_ip_min_max(), (nidrange->nr_all == 1) means this nid range
is a full IP address range(*.*.*.*). In this case, we don't need
to compare it to any other nid range, but set min_nid to 0.0.0.0
and max_nid to 255.255.255.255 directly.

Also, test_10d is added to sanity-sec.sh to verify this patch and
some code cleanup is done for jt_nodemap_add/del_range().

Change-Id: I72c546b060f9e123204a566a3bd373b4f017502d
Signed-off-by: Emoly Liu <emoly.liu@intel.com>
Reviewed-on: https://review.whamcloud.com/31684


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: Fan Yong <fan.yong@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

In kiblnd_peer_connect_failed() replace a backwards list_add() and
list_del() with list_splice_init().

Signed-off-by: John L. Hammond <john.hammond@intel.com>
Change-Id: Ib00d5d911d1070b6c8b49f14a2c7fc3552da553c
Reviewed-on: https://review.whamcloud.com/31643


Tested-by: Jenkins
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

This reverts commit c3894ff8. Dropping
early receives caused pings to be ignored and interacted badly with
dynamic discovery.

Signed-off-by: John L. Hammond <john.hammond@intel.com>
Change-Id: I99a87a8f58ea67c59d5e85b964295472c2e15de4
Reviewed-on: https://review.whamcloud.com/31675


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Discovery protocol sends a ping (GET) to the peer and expects a
REPLY back with the interface information. Discovery uses the
DEFAULT_PEER_TIMEOUT which 180s. This could lead to extended delay
during mounting if the OSTs are down or if the ping fails for
any reason.

This patch adds a module parameter lnet_transaction_timeout which
defaults to 5 seconds. lnet_transaction_timeout is used for the
discovery timeout.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: Ida1e19f55552b24e83c8094aa88a37c2748126cf
Reviewed-on: https://review.whamcloud.com/31663


Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

In lnet/utils/lnetconfig/cyaml.c, for free_node()
check first for NULL pointer before dereferencing it.

Issue found in Static analysis

Change-Id: I6298f0f09175b6fd210db5717d44d050b1cb9d8d
Signed-off-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-on: https://review.whamcloud.com/31551


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Supress few false warnings with a compiler option when
building the lustre kernel modules. A few other fixes
to make lustre buildable on Fedora.

Change-Id: If14d226e5d92ae9ce54e216d032df94d9398654e
Signed-off-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-on: https://review.whamcloud.com/28810


Tested-by: Jenkins
Reviewed-by: John L. Hammond <john.hammond@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

With the migration to 64 bit second time some calls to lnet_notify
did not get updated to use time64_t. Update those calling points.
Also for the ioctl IOC_LIBCFS_NOTIFY_ROUTER we pass in the number
of seconds since the epoch. We subtract the current epoch time but
we missed adding in the current number of seconds since booting
since the lnet ping code expects the seconds since boot to be used.

Change-Id: I5a92df08cdaf3b747fd17721a92038df05669a81
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/31339


Reviewed-by: Doug Oucharek <dougso@me.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

This patch fixes the contiguous range check to allow the addition of
multiple "full" ([0-255]) ranges. As part of this change,
is_contiguous and find_min_max are combined as they were always
called together and the logic is fairly similar. This also removes
the multiple range expression support, since it was broken.

Also, sanity-sec.sh test_10c is added to verify this patch.

Signed-off-by: Kit Westneat <kit.westneat@gmail.com>
Signed-off-by: Emoly Liu <emoly.liu@intel.com>
Change-Id: I3c49a077039327fcbde87196f82db140f67a74d0
Reviewed-on: https://review.whamcloud.com/24397


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

list_add was being used erroneously. The logic should be to move
the txs on ibp_tx_queue on a local list which is then processed.
The code, however, did the reverse, which would result in the
pending txs not processed and thus dropped silently. This in turn
would lead to peers reference counts at the LNet layer not
decremented since lnet_finalize() might not be called for a message.

Initialize local list and use list_splice_init() to move
transmits on the ibp_tx_queue to the local list.

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I6b36f709db2c89e53e0b3354883a8a1b1052a1dd
Reviewed-on: https://review.whamcloud.com/31374


Tested-by: Jenkins
Reviewed-by: Doug Oucharek <dougso@me.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Remove remaining libcfs time wrappers from lnet selftest. Migrate
crp_stamp to nanoseconds and both timestamps nd_stamp, sn_start to
ktime. The move away from jiffies which can vary on platforms to
something that is the consistent on every node. This will ensure
that the reported results to the user will always be correct.

Test-Parameters: trivial testlist=lnet-selftest

Change-Id: Id8d1b195f690c69635de60dd9b501f6d97f90f4d
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/31041


Tested-by: Jenkins
Reviewed-by: Doug Oucharek <dougso@me.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Add a "lfs setstripe --yaml=<yaml_template> <lustre_file_or_dir>"
usage to set stripe using stripe info from a YAML template file.

The YAML template file can be get from
$ lfs getstripe --yaml <lustre_file_or_dir>
and user can manually edit it to tweak stripe options.

This patch fixes two cyaml issues:
1. a YAML_BLOCK_ENTRY_TOKEN can follow a YAML_VALUE_TOKEN.
2. free_node() has memory leak, it needs to free
   cYAML::cy_valuestring and cYAML::cy_string if possible.

Test-Parameters: testlist=sanity-pfl,sanity-flr
Signed-off-by: Bobi Jam <bobijam.xu@intel.com>
Change-Id: I78149bb011fbc03387cbe3d057eb030550dd75ae
Reviewed-on: https://review.whamcloud.com/26860


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Emoly Liu <emoly.liu@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

To avoid confusion this fix moved the freeing a struct kib_conn outside of
the function kiblnd_destroy_conn().

Change-Id: Iae28802f5d319570064a504feb14dffd13a22b84
Signed-off-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-on: https://review.whamcloud.com/31273


Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

early RX added to the o2ib lnd as attempt to reordering problem
handling, When messages have arrived before actual connection sets.
But it code can fill all incoming queue and normal connect will not
processed.

Cray-bug-id: MRP-4638
Signed-off-by: Alexey Lyashkov <c17817@cray.com>
Change-Id: I2efc73534a20c4628ed462ee5055c901dbf44278
Reviewed-on: https://review.whamcloud.com/30254


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The 4.13 kernel renames wait_queue_t to wait_queue_entry_t.

Add a probe and handle rename across the code base and have
a define to translate to the new name when indicated.

Test-Parameters: trivial

Change-Id: I8f0f5ec4d02ccb270acb72ccffe13f0ecf6bd2f7
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-on: https://review.whamcloud.com/31153


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Ben Evans <bevans@cray.com>
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>

Remove remaining libcfs time wrappers from ko2iblnd. Also fix bug
in ksocklnd to use cfs_time_seconds for calling schedule_timeout
instead of cfs_durations_sec. That was the opposite of the
conversion we needed. the remaining jiffy use it moved to
time64_t.

Change-Id: I5847d7260ac8a9be1b165423adb7b8e9a53998d2
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/31042


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

There is a performance drop when using IB_MR_TYPE_SG_GAPS. To
mitigate this, we added a module parameter, use_fastreg_gaps, which
defaults to 0. When allocating the memory region if this parameter
is set to 1 and the hw has gaps support then use it and output
a warning that performance may drop. Otherwise always use
IB_MR_TYPE_MEM_REG.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I08a8b72756b9b5b5bcb391bf3e979f6d28eb5cbb
Reviewed-on: https://review.whamcloud.com/30749


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

If the net information is not provided while adding
NI, it results in an LBUG. Remove the LASSERT on
userspace input and handle it gracefully.

Test-Parameters: trivial
Change-Id: I9d2b6f94cb35e94bc81d5c52936d32cbf833e597
Signed-off-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-on: https://review.whamcloud.com/31100


Tested-by: Jenkins
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The level of verbosity is high when exporting the YAML configuration
for the purposes of storing it to reconfigure a node. This patch
eliminates the unnecessary YAML lines which are not needed when
reconfiguring a node, such as statistics, status, etc.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: Ie57c761415cfb0ceee8b2dbc0b293e85ae415685
Reviewed-on: https://review.whamcloud.com/30845


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Move debug print within the cpt lock to keep
peer access safe.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: Ic37ff0973367b3eb9cbc0059ffee9c31ecf98c34
Reviewed-on: https://review.whamcloud.com/28771


Tested-by: Jenkins
Reviewed-by: John L. Hammond <john.hammond@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Upon async PUT request received from peer already in discovery
process, lnet_peer_push_event() was not handling the case where
peer could be on working/ln_dc_working queue. This could lead
for peer not to be re-dsicovered as expected, but left on
working queue and to be finally timed-out.

Also ensure that peer will not be put back on request queue by
event handler if discovery is already completed.

Signed-off-by: Bruno Faccini <bruno.faccini@intel.com>
Change-Id: Ic74a313c00edc1b8fdd14794d2c88411d12e0979
Reviewed-on: https://review.whamcloud.com/30147


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Starting with gcc version 7 several platforms have enabled new
flags to report potential problems when compling code. For lustre
much of the reported problems deal with potential buffer overruns.
Also we have unused data structures and are not properly
initializing some data structures.

Change-Id: I10243ea88f2c726032d179febdbf26f28de13715
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/30376


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Recently, a change has been merged to "lnet/klnds/o2iblnd/o2iblnd.c" which
introduces the usage of IB_DEVICE_SG_GAPS_REG and IB_MR_TYPE_SG_GAPS.

Unfortunately, these symbols are not available in the 4.4.0 Kernels as used
by Ubuntu 14/16.

Additionally, there seems to be general warning against their use:
 - https://patchwork.kernel.org/patch/9573483/
 - https://lkml.org/lkml/2017/3/13/206



 Also, there is a related performance issue as reported in LU-10394.

The solution is to create a preprocessor guard around their use, so that
Kernels lacking these symbols will not use them and revert to using the older
IB_MR_TYPE_MEM_REG, instead.

Test-Parameters: trivial
Signed-off-by: Martin Schroeder <martin.h.schroeder@intel.com>
Change-Id: Ie835d6e04f3859634ba508c24dff1f27f1b24cf6
Reviewed-on: https://review.whamcloud.com/30893


Tested-by: Jenkins
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Migrate away from jiffies for the pinger to time_64_t to one make
it clear its for time keeping and secondly to ensure the behavior
is consistent across any platform. Besides the lnet pinger code
move the lnet dely injection code to time64_t as well.

Change-Id: If363523893fc1dcce4eaa866501946edd6558751
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/30658


Tested-by: Jenkins
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

GNI interfaces don't have IP addresses so when configuring GNI
interfaces there is no point of trying to query the ip. There is
also only one GNI interface, therefore the net configuration
command shouldn't enforce an interface name.

This patch also adds more descriptive error commands. It also allows
deleting an entire network without having to specify an interface.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I549647675fe5530db7d86272a7dc79892117847d
Reviewed-on: https://review.whamcloud.com/30829


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Throttle the transmits based on the negotiated conn queue depth
to ensure we keep the number of outstanding transmits below the
negotiated queue depth.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I27190364904d6c79c0cd6d382228f8b8d2b11ba0
Reviewed-on: https://review.whamcloud.com/30751


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Lustre uses libcfs.h as a the header to include all headers. This
approach has drawbacks like colliding with MOFED compat headers
that do the same thing. This patch is the first step to unwind
including libcfs.h everywhere. This starts with eliminating
linux/libcfs.h.

Test-Parameters: trivial

Change-Id: Id2040d4295c16135561c8251e160cb2117ee21b8
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/30706


Reviewed-by: Doug Oucharek <dougso@me.com>
Tested-by: Jenkins
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>